FlowKV: A Disaggregated Inference Framework with Low-Latency KV Cache Transfer and Load-Aware Scheduling

Li, Weiqing, Jiang, Guochao, Ding, Xiangyong · arXiv · 2025